Self-Recovery E-evidence Image Watermarking Algorithm Based on Binary features of Important Blocks

Existing self-recovery watermarking algorithms are generally designed for natural images, which have a large watermark embedding capacity and cannot clearly recover the text information. To deal with this, a self-recovery watermarking algorithm for E-evidence image authentication was proposed, which can reach a balance between watermark embedding capacity and recovery quality. Firstly, according to the characteristics of an E-evidence image, all image blocks in the test E-evidence image were categorized into three types:important block, unimportant one and blank one. Then, to ensure the correctness of the type code, the type code is permutated and encoded by RS coding to generate a partial-recovery watermark that is randomly embedded in all blocks. At the same time, another recovery watermark is generated by coding the binary features of important blocks and randomly embedded in the least significant bit of unimportant or blank blocks based on secret key. At last, the performances of the proposed scheme and the alterable-capacity watermarking algorithms were compared by using seven E-evidence images. Results show that the watermark embedding capacity of the proposed scheme is reduced to 0.64 bit per pixel, compared with two literature results of 1.73 and 2.99, and the peak signal to noise ratio (PSNR) of the watermarked E-evidence image by the proposed scheme is about 8 dB and 23 dB higher than those obtained from literature, while the performance of tamper detection and tamper recovery is not worsened. This is mainly because the proposed scheme adopts classification code and only encodes the important blocks.